Transforming Influencer Discovery with AI-Powered Semantic Search

The Challenge

GOAT APAC analyzes large-scale social media posts every month, a crucial asset for influencer discovery, emphasizing the need for fast and efficient search. However, their previous search capabilities were severely limited:

• Keyword Shackles: Search functionality was restricted to simple keyword matching within post captions and hashtags.
• Visual Blind Spot: The rich context and data contained within images and videos, the core of influencer content, was completely inaccessible to search.
• Inefficient Matching: Identifying the perfect influencer for a specific brand brief required countless hours of manual, inefficient searching, leading to missed opportunities and suboptimal pairings.

 

GOAT needed an intelligent solution to unlock the meaning, context, and visual data within the social media posts.

The Solution

Datatonic partnered with GOAT to architect and implement a sophisticated, scalable AI solution to unlock rapid analysis of social media posts at scale. This solution fundamentally upgraded the analysis as it introduces the capability of semantic search. We developed two core, interoperable services hosted on Cloud Run and orchestrated by Workflows:

 

1. The Media Analyser (Powered by Gemini)

This service analyzes all visual media (images and videos) to understand context and intent. This includes branding and logo detection, identifying direct brand references, competitor logos, and product placements within the visual media, as well as context generation: determining the activity, setting, and purpose of the content (e.g., “travel vlog,” “cooking demonstration,” “fashion haul”).

 

2. The Embeddings Generator

This service utilizes advanced machine learning to translate text and metadata into numerical vector representations (embeddings), enabling Semantic Search.

Instead of matching keywords, the system now understands the intent and context of a search query, connecting concepts and ideas to deliver highly relevant results.

Datatonic leveraged several Google Cloud services for this project, including Gemini for media analysis and context extraction, BigQuery for high-throughput analysis and semantic matching at scale, Workflows for pipeline orchestration, and Cloud Run for scalable, managed hosting of services.

 

Functionality + Implementation

Datatonic delivered the core AI capability and a test interface. GOAT’s engineering team successfully integrated the new embeddings API and enriched metadata directly into their internal, proprietary Brand.AI platform, ensuring immediate adoption and scalability.

The new capabilities delivered included:

• Natural Language Search: Internal planners can now search using complex, contextual phrases (e.g., “Influencers promoting sustainable swimwear in Bali last summer”).
• Multi-Lingual Search: Crucial for the APAC region, the system accurately processes and matches queries across multiple local languages.
• Visual-Text Correlation: The ability to search both content assets and the creators themselves.
• Enhanced Clarity: A ranking score is provided alongside the search results, giving users full clarity on how content was matched to their criteria.
• Brand Safety Extraction: Automated identification of competing brands or flagged content for sophisticated risk management.

 

Our impact

By moving beyond simple keyword matching, GOAT gained an unprecedented level of granularity and efficiency in its influencer discovery process.

• Search Method: The search method transitioned from simple Keyword Matching to Contextual Semantic Search using Gemini and Embeddings, resulting in 100% contextual relevance.
• Content Unlocked: Content that was previously restricted to Captions and Hashtags only now includes Images and Videos, unlocking understanding of large-scale social media data.
• Reduced Search Time: Manual search time was reduced from hours to near-Instant after the implementation.
• New Discovery Use Cases:
  Cross-Brand competitor analysis – Through automated branding and logo detection, GOAT can now identify direct brand references and competitor placements within visual media that were previously “blind spots.” This allows for an unprecedented ability to compare content across brands, enabling planners to analyse a competitor’s influencer footprint, visual strategy, and the specific settings where rival products are being featured.
  In-Depth Brand Discovery – The platform can now conduct a holistic, 360-degree analysis of a brand’s presence by correlating text metadata with visual context. Planners can use natural language queries to discover how a brand is truly being represented by creators – identifying not just who is tagging the brand, but the intent and “vibe” of the content, such as identifying influencers promoting sustainable themes or specific product placements in diverse markets.

The implementation of Gemini and vector embeddings not only saved significant manual effort but also created a more intelligent, future-proof platform, enabling GOAT to deliver precise, high-value influencer recommendations across the diverse and rapidly evolving APAC markets but it has increased their search capabilities and the level of accuracy.